We present simulations suggesting that it is possible to minimize the systematic errors of differential absorption lidar (DIAL) measurements caused by the Rayleigh-Doppler effect by selecting an online frequency close to one of the inflection points on either side of the absorption line. Thus, it seems advantageous to select an absorption line of suitable cross section at these points on the line slopes rather than at the peak. First, we extend the classical simulation study of Ansmann (1985) for another water vapor absorption line but again with the online frequency at the line peak. As expected, we also found large systematic errors of more than 40% at the edges of aerosol layers and clouds. Second, we simulate the systematic errors for other online frequencies away from the peak for the same input profile. The results demonstrate that the errors vanish close to the inflection points. Since both the shape of the absorption lines and the width of the broadened backscatter signal depend on the atmospheric conditions, these optimum frequencies vary slightly with height and climatology. Third, we calculate the errors for a typical aerosol profile of the planetary boundary layer obtained from lidar measurements. With this case, we discuss how to select practically the online frequency so that the errors are minimized for all heights of interest. We found that the error reduces from 20 to < 1% at the top of the planetary boundary layer while, at the same time, the error reduces from 6 to 2% in 5 km.